INFECTION AND IMMUNITY, JUlY 1986, p. 135-1400019-9567/86/070135-06$02.00/0Copyright © 1986, American Society for Microbiology

Two Toxin-Converting Phages from Escherichia coli 0157:H7 Strain933 Encode Antigenically Distinct Toxins with Similar

Biologic ActivitiesNANCY A. STROCKBINE,' LILIAN R. M. MARQUES,' JOHN W. NEWLAND,' H. WILLIAMS SMITH,2

RANDALL K. HOLMES,' AND ALISON D. O'BRIENl*

Department of Microbiology, Uniformed Services University of the Health Sciences, Bethesda, Maryland 208144799,'and Houghton Poultry Research Station, Huntingdon, Cambridgeshire, PEJ7 2DA United Kingdom2

Received 17 January 1986/Accepted 31 March 1986

Escherichia coli 0157:H7 strain 933 contains two distinct toxin-converting phages (933J and 933W). Thebiologic activities and antigenic relationship between the toxins produced by 933J and 933W lysogens of E. coliK-12, as well as the homology of the genes that encode the two toxins, were examined in this study. The 933Jand 933W toxins, like Shiga toxin produced by Shigella dysenteriae type 1, were cytotoxic for the same cell lines,caused paralysis and death in mice, and caused fluid accumulation in rabbit ileal segments. The cytotoxicactivity of 933J toxin for HeLa cells was neutralized by anti-Shiga toxin, whereas the activity of 933W toxin wasnot neutralized by this antiserum. In contrast, an antiserum prepared against E. coli K-12(933W) neutralized933W toxin but not 933J toxin or Shiga toxin. For E. coli 933, most of the cell-associated cytotoxin was

neutralized by anti-Shiga toxin, whereas most of the extracellular cytotoxin was neutralized by anti-933Wtoxin. However, a mixture of these antisera indicated the presence of both toxins in cell lysates and culturesupernatants. Among 50 elevated cytotoxin-producing strains of E. coli, we identified 11 strains isolated fromcases of diarrhea, hemorrhagic colitis, or hemolytic uremic syndrome that produced cell-associated cytotoxinswhich were neutralized by the 933W antitoxin. Southern hybridization studies showed that the cloned toxinstructural genes from phage 933J hybridized with DNA from phage 933W under conditions estimated to allowno more than 26% base-pair mismatch. These findings indicate that E. coli produces two genetically related butantigenically distinct cytotoxins with similar biologic activities which we propose to name Shiga-like toxins Iand II. Strains of E. coli that produce elevated levels of Shiga-like toxin I or Shiga-like toxin II, or both, havebeen associated with the clinical syndromes of diarrhea, hemorrhagic colitis, and hemolytic uremic syndrome.

Escherichia coli serotype 0157:H7 was incriminated as

the agent responsible for two outbreaks of hemorrhagiccolitis that occurred in the United States during 1982 (23,26). The E. coli 0157:H7 strains isolated during theseoutbreaks did not produce either heat-labile or heat-stableenterotoxin, nor were they enteroinvasive (26). However,two isolates from patients and an isolate (strain 933) fromground beef, which was implicated as the vehicle in theseoutbreaks, produced high levels of cell-associated and cell-free cytotoxin active on both HeLa and Vero cells (A. D.O'Brien, T. A. Lively, M. E. Chen, S. W. Rothman, andS. B. Formal, Letter, Lancet i:702, 1983). In Canada, strainsof E. coli 0157:H7 have been isolated from patients withhemorrhagic colitis (20; W. M. Johnson, H. Lior, and G. S.Bezanson, Letter, Lancet i:76, 1983) and hemolytic uremicsyndrome (6; M. A. Karmali, B. T. Steele, M. Petric, and C.Lim, Letter, Lancet i:619-620, 1983), and these strains havebeen shown to produce a cytotoxin active on Vero cells.Toxin from E. coli 933 was purified to apparent homogeneityand shown to have biologic activities similar to those ofShigella dysenteriae type 1 (Shiga) toxin: both toxins were

cytotoxic for selected cell lines in vitro, paralytic and lethalfor mice, and enterotoxic in ligated rabbit ileal segments (8;A. D. O'Brien, T. A. Lively, T. W. Chang, and S. W.Gorbach, Letter, Lancet ii:573, 1983).

Production of elevated levels of a HeLa or Vero cellcytotoxin by E. coli is associated with temperate bacterio-phages (17, 24; S. M. Scotland, H. R. Smith, G. A. Will-

* Corresponding author.

shaw, and B. Rowe, Letter, Lancet ii:216, 1983; H. R.Smith, N. P. Day, S. M. Scotland, R. J. Gross, and B.Rowe, Letter, Lancet i:1242-1243, 1984). E. coli 933 harborstwo distinct toxin-converting phages designated 933J and933W (17). The cytotoxin produced in large amounts by E.coli K-12 lysogenized with phage 933J can be neutralized byantiserum against Shiga toxin (17). The structural genes forthis toxin were cloned from phage 933J DNA and shown bySouthern hybridization to have homology with DNA from S.dysenteriae type 1 and Shigellaflexneri (13). Recent studiesin our laboratory demonstrated that the cytotoxin producedby E. coli K-12(933W) was not neutralized by anti-Shigatoxin, in contrast with our original report on 933W toxin (17).The purposes of this investigation were to compare thebiologic activities and antigenic specificities of the 933J and933W toxins and to determine whether their structural genesare homologous.(A preliminary report of this work was presented at the

Twenty-first Joint Conference on Cholera sponsored by theU.S.-Japan Cooperative Medical Sciences Program, Octo-ber 21-23, 1985, at the National Institutes of Health, Be-thesda, Maryland.)

MATERIALS AND METHODS

Bacterial strains. E. coli 0157:H7 strain 933 was obtainedfrom G. K. Morris, Centers for Disease Control, Atlanta,Ga. E. coli C600(933J) and E. coli C600(933W) were con-

structed by lysogenizing E. coli K-12 strain C600 with phage933J or phage 933W (17). E. coli H30 (026:H11) was

obtained from J. Konowalchuk, Bureau of Microbial Haz-

135

Vol. 53, No. 1

on June 9, 2018 by guesthttp://iai.asm

.org/D

ownloaded from

136 STROCKBINE ET AL.

ards, Ottawa, Canada. S. dysenteriae type 1 strain 60R wasobtained from S. B. Formal, Walter Reed Army Institute ofResearch, Washington, D.C. Clinical isolates of E. coli wereobtained from M. A. Karmali, Hospital for Sick Children,Toronto, Canada; from D. Sherwood, Moredun ResearchInstitute, Edinburgh, Scotland; from J. G. Wells, Centersfor Disease Control, Atlanta, Ga.; and from B. Rowe,Central Public Health Laboratory, London, United King-dom. Strains were routinely stored at -70°C in Penassaybroth (Difco Laboratories, Detroit, Mich.) supplementedwith 15% glycerol.

Toxins. Bacteria were grown for 48 h in Chelex-treatedsyncase medium as previously described (16). Culture super-natants were collected and bacterial lysates were preparedeither by the French press method (16) or by sonication (14).

Polyclonal rabbit antisera. Antibody to purified Shiga toxinwas raised in rabbits according to published procedures (15).Antibody to crude 933W toxin was raised in rabbits by aseries of 13 intramuscular inoculations of extracellular cul-ture fluid (0.2 ml) or cell extracts (0.05 to 0.2 ml) preparedfrom E. coli C600(933W) as previously described (24), mixed1:1 with incomplete Freund adjuvant, and administered atirregular intervals over a period of 53 days.

Cytotoxicity assays. Filter-sterilized culture supernatantsand bacterial lysates were tested for cytotoxicity (4) onHeLa cells, Vero cells, Chinese hamster ovary cells (CHO),and mouse Y-1 adrenal cells. HeLa cells were grown inEagle minimal essential medium (Flow Laboratories, Inc.,McLean, Va.); Vero and Y-1 adrenal cells were maintainedin RPMI 1640 medium (Flow Laboratories); and CHO cellswere maintained in alpha Eagle minimum essential medium(Flow Laboratories). Media were supplemented with 10%(vol/vol) fetal calf serum, 0.8 mM glutamine, 500 U ofpenicillin G per ml, and 500 p.g of streptomycin per ml.Microtiter plates (96-well; Costar, Cambridge, Mass.) wereinoculated with approximately 10,000 cells per well (100 VLl)and incubated overnight at 37°C in the presence of 5% CO2.Serial dilutions of culture supernatants or bacterial lysateswere made in cell culture medium. Samples containing 100VLI of each dilution were added to wells of the microtiterplates, and incubation was continued at 37°C in 5% CO2.Vero cells were examined daily for 4 days by light micros-copy for cytotoxicity, whereas the HeLa, CHO, and Y-1adrenal cells were examined after 24 h. In each case, the 50%cytotoxic dose (CD50) corresponded to the amount of toxinrequired to kill 50% of the cells in a well.

Neutralization of cytotoxicity. Toxin from culture superna-tants or sonic lysates was serially diluted (twofold for lowtoxin producers or 10-fold for moderate or high toxin pro-ducers) in cell culture medium. A sample of each toxindilution (100 p.l) was mixed with an equal volume of one ofthe antibodies described below and incubated at 37°C for 1 hfollowed by overnight incubation at 4°C. Samples (100 pJ) ofeach toxin-serum mixture were added to individual wells ofmicrotiter plates containing HeLa cells, and the cells wereexamined for cytotoxic effects after incubation overnight at37°C in 5% CO2. The antibodies used for neutralization testsincluded undiluted hybridoma culture supernatant contain-ing monoclonal antibody (MAb) 13C4 specific for the Bsubunit of Shiga-like toxin (25), MAb 32D3 specific for the Bsubunit of cholera toxin (5) as a negative control, and 1:50dilutions in cell culture medium of rabbit antiserum againstpurified Shiga toxin, rabbit antiserum against crude 933Wtoxin, or normal rabbit serum as a negative control. Lysatesof E. coli H30 and S. dysenteriae type 1 strain 60R wereincluded in the neutralization assays as toxin controls to

demonstrate the neutralizing activities of MAb 13C4 and therabbit antiserum against Shiga toxin. A sample was consid-ered to contain neutralizable cytotoxin if the dose requiredto kill 50% of the HeLa cells in the presence of antitoxinantibody was at least fourfold greater for samples dilutedtwofold or tenfold greater for samples diluted tenfold thanthe amount of toxin required for equivalent cytotoxicity inthe presence of control antibody. These assays for Shiga-liketoxins are similar in principle to the assays used to determineL+ and Lr doses of diphtheria toxin (28).Mouse lethality. The lethality of crude French press ly-

sates from E. coli 933, E. coli C600, E. coli C600(933J), andE. coli C600(933W) was tested according to published pro-cedures (18). The 50% lethal dose (LD50) of each toxin wasdetermined in the following way. Groups of five female (6 to8 weeks old) CD-1 mice (Charles River Breeding Laborato-ries, Inc., Kingston, N.Y.) were inoculated intraperitoneallywith various doses of bacterial lysates diluted in phosphate-buffered saline that contained 0.1% (wt/vol) gelatin. Themice were observed daily over a 10-day period for hind-legparalysis and death. The LD50s were calculated by themethod of Reed and Muench (21).

Enterotoxicity assay. French press lysates of E. coli 933, E.coli C600, E. coli C600(933J), and E. coli C600(933W) weretested for enterotoxic activity in ligated rabbit ileal segmentsas previously described (3). One milliliter of bacterial lysate(undiluted) was inoculated per segment. Each sample wastested in eight ligated segments of different rabbits. A samplewas considered to contain enterotoxin if the average of thevolume-to-length ratios of the eight inoculated segments was-1.0 ml/cm.

Restriction endonuclease digestion and hybridization anal-ysis of 933W phage DNA. DNA from phages 933W and 933Jwas prepared as previously described (17) and digested withBamHI, Salil, or EcoRI (International Biotechnologies, Inc.,New Haven, Conn.) according to the recommendations ofthe manufacturer. DNA fragments were separated electro-phoretically on 0.7% (wt/vol) agarose gels, blotted ontonitrocellulose paper, and probed as previously described (2,12). Probe pJN25 contains a 3.0-kilobase insert from phage933J that encodes the 933J cytotoxin (13). Approximatelyhalf of the 3.0-kilobase insert from phage 933J is estimated tocode for toxin, while the other half represents phage se-quences. The pJN25 probe was radiolabeled by nick trans-lation using 32P-labeled dCTP and a commercially availablekit (Bethesda Research Laboratories, Inc., Gaithersburg,Md.). The hybridization with the 32P-labeled pJN25 probewas performed in 5x SSPE (0.9 M NaCl, 50 mMNaH2PO4* 2H2O, 5 mM disodium EDTA; pH 7.4) at 50°C or65°C overnight. Washing was done at room temperature ifthe hybridization was performed at 50°C or at 50°C if thehybridization was performed at 65°C. As a control, thedigested 933W phage DNA was also probed with 32P-labeledpBR328 (the vector plasmid contained in pJN25) underconditions allowing approximately 36% mismatch. The blotswere exposed to Kodak X-AR film (Eastman Kodak Co.,Rochester, N.Y.) in the presence of intensifying screens.

RESULTS

Biologic activities of the 933J and 933W toxins. As shown inTable 1, lysates of E. coli 933, E. coli C600, E. coliC600(933J), and E. coli C600(933W) had the same specificityfor killing cultured cells. Lysates from all four strains killedHeLa and Vero cells but did not affect Chinese hamsterovary cells or Y-1 adrenal cells. Culture supernatants of E.

INFECT. IMMUN.

on June 9, 2018 by guesthttp://iai.asm

.org/D

ownloaded from

ANTIGENIC VARIATION AMONG SHIGA-LIKE TOXINS OF E. COLI 137

TABLE 1. Comparison of the biologic activities of Shiga-like toxins in cell extracts of E. colia

E. coli strain Cytotoxicity for: Paralysis and death Enterotoxicity in ligatedHeLa cells Vero cells CHO cells Y-1 cells in mice rabbit ileal segments

933 + (high) + (high) - - + +C600 + (low) + (low) - - _C600(933J) + (high) + (high) - - + +C600(933W) + (moderate) + (moderate) - - + +

a Symbols: +, presence of biologic activity; -, absence of biologic activity. The level of cytotoxicity (Marques et al., in press) is given in parentheses: high,105 to 108 CD50ml in sonic lysates; moderate, 103 to 104 CD50/ml in sonic lysates; low, 2 x 101 to 6 x 102 CD5o/ml in sonic lysates.

coli 933, E. coli C600(933J), and E. coli C600(933W) werealso selectively cytotoxic for HeLa cells (Table 2) and Verocells (not shown), but the culture supernatant of E. coli C600contained no demonstrable cytotoxic activity. This is thesame pattern of cytotoxicity that has been reported for Shigatoxin from S. dysenteriae type 1 (9, 19).The amounts of cytotoxin present in extracts of E. coli

933, E. coli C600, E. coli C600(933J), and E. coli C600(933W)differed greatly (Tables 1 and 2). Typically, sonic lysates ofE. coli 933 were 10,000- to 100,000-fold more toxic thanlysates from E. coli C600. Lysates from E. coli C600(933J)were approximately 1,000-fold more toxic than lysates fromE. coli C600(933W). In contrast, the toxicity of sonic lysatesof E. coli C600 did not exceed 100 CD5W/ml and was muchlower than the toxicity of lysates from C600(933J) orC600(933W). The cytotoxicity of extracts prepared by theFrench press method was typically greater than thecytotoxicity of sonic lysates by about 10-fold.French press lysates of E. coli 933, E. coli C600(933J), and

E. coli C600(933W) were capable of paralyzing and killingmice. The lysate of E. coli C600 did not cause these effects(Table 1), presumably because the amount of cytotoxin in E.coli C600 lysates was below the level necessary for mouselethality (O'Brien, et al., Lancet ii:573, 1983). In two exper-iments, the number of LD5so per milligrams of protein for E.coli C600(933J) was 160 to 170, whereas for E. coliC600(933W) the, number of LD5so per milligrams of proteinwas 8 to 14. The number of CD50s corresponding to one

LD5os was 40- to 70-fold greater for 933J toxin than for the933W toxin.

Lysates of E. coli 933, E. coli C600(933J), and E. coli

C600(933W) also had enterotoxic activity in the rabbit li-gated ileal segment assay. The E. coli C600 lysate did notcause fluid accumulation (Table 1), but because the numberof CD50s contained in 1 ml of C600 lysate was also expectedto be below the threshold for enterotoxicity (O'Brien et al.,Lancet ii:573, 1983). Dose-response curves were not per-formed in ileal segments because the number of rabbitsavailable was limited. The data in Tables 1 and 2 demon-strate that the 933J and 933W toxins have the same kinds ofbiologic activities. The greater lethality relative tocytotoxicity for 933W toxin was reproducible and was themost striking difference between the biologic activities ofthese two toxins.

Antigenic relationships between the 933J and 933W toxins.Cross-neutralization studies were done to assess the anti-genic relationship between the 933J and 933W toxins (Table2). In these assays, a constant amount of antibody wasmixed with various amounts of lysate or supernatant. Underthese circumstances, cytotoxicity is detected if the amountof homologous cytotoxin exceeds the neutralizing capacityof the antitoxin, or heterologous (non-neutralizable) cyto-toxin is present, or both. To distinguish among these alter-natives, samples of lysates or supematants were also treatedwith mixtures of antisera capable of neutralizing the 933J and933W toxins. The cytotoxic activities in the lysate andculture supernatant of E. coli C600(933J) were neutralized bymonoclonal antibody against the B subunit of Shiga-liketoxin (MAb 13C4) or by anti-Shiga toxin, but they were notneutralized by anti-933W toxin. In contrast, the cell-associated and cell-free cytotoxins from E. coli C600(933W)were neutralized by anti-933W toxin, but they were not

TABLE 2. HeLa cytotoxicity of culture supernatants and lysates from E. coli strains in the presence or absence of antisera againstShiga and Shiga-like toxinsa

No. of CD50 required for cytotoxicity in the presence of:

E. coli strain CD50/ml a-CT MAb a-SLT Rabbit Rabbit Rabbit

32D3 NRS MAb 13C4 at-Shiga oa-933W a-933W

C600(933J)Lysate 10,000,000 1 1 1,000 10,000 1 10,000Supernatant 100,000 1 1 1,000 1,000 1 1,000

C600(933W)Lysate 10,000 1 1 1 1 100 100Supernatant 10,000 1 1 1 1 100 100

933Lysate 10,000,000 1 1 100 100 1 10,000Supernatant 1,000,000 1 1 1 1 100 10,000a Rabbit sera were used at dilutions of 1:50, and the monoclonal antibodies were used as undiluted hybridoma culture supernatant. Neutralization controls

included lysates from S. dysenteriae 1 strain 60R and E. coli H30. The activity in these lysates was neutralized by anti-Shiga toxin and a-Shiga-like (a-SLT) MAbbut not by a-cholera toxin (a-CT) MAb, normal rabbit serum (NRS) or rabbit anti-933W. The lysates were prepared from suspensions of bacterial cells that hadbeen concentrated approximately 50-fold with respect to the original cultures from which the supernatants were obtained, and the data shown are representativeof the results of many experiments. The titer of a sonic lysate of E. coli C600 was 100 CD50/ml. Culture supernatants of E. coli C600 were not cytotoxic.

VOL. 53, 1986

on June 9, 2018 by guesthttp://iai.asm

.org/D

ownloaded from

138 STROCKBINE ET AL.

neutralized either by MAb 13C4 or by anti-Shiga toxin. Thislack of cross-neutralization demonstrates that the phage-encoded 933J and 933W toxins are antigenically distinct.The cytotoxicity of extracts of E. coli C600 was usually

neutralized by MAb 13C4 and by rabbit anti-Shiga toxin(data not shown), confirming previous reports (16, 25), butthe rabbit anti-933W toxin did not neutralize the E. coli C600cytotoxin. Occasionally either no activity could be detectedwith E. coli C600 lysates or they exhibited low titers ofnonspecific cytotoxicity that could not be neutralized eitherwith anti-Shiga toxin or with anti-933W toxin.The lysate of E. coli 933 contained a high level of cytotoxic

activity (Table 2) which was neutralized by MAb 13C4 oranti-Shiga toxin, but which was not neutralized by rabbitanti-933W toxin. However, a mixture of rabbit anti-Shigatoxin and rabbit anti-933W toxin neutralized the toxicity ofthe E. coli 933 lysate to a greater extent than anti-Shiga toxinalone, suggesting that a second toxin was present at a lowerlevel. In contrast, the mixture of antitoxins was no moreeffective against the toxin from C600(933J), or C600(933W)than was the single, active antitoxin in the mixture. Thesecontrol experiments demonstrated that anti-Shiga toxin andanti-933W antitoxins did not act additively or synergisticallyto enhance the neutralization of 933J or 933W cytotoxin.Therefore, the results obtained with lysates of strain 933demonstrated the presence of two distinct cytotoxins whichwere antigenically related to the toxins encoded by phages933J and 933W.

Culture supernatants of E. coli 933 also contained 933J and933W toxin, but their relative titers varied from one super-natant to another. This variability has led to inconsistentneutralization results for the extracellular cytotoxicity withanti-Shiga toxin. Our findings with this antiserum and culturesupernatants from strain 933 have ranged from completeneutralization (O'Brien et al., Lancet i:702, 1983) to partialor no demonstrable neutralization. A second problem thathas also led to inconsistent neutralization results is the lossof 933W activity in some cultures of 933 left at 4°C on agarplates for more than 3 months.

In the experiment presented in Tab-le 2, the relative titersof 933J and 933W toxin in the culture supernatant of strain933 were the reverse of those seen in cell lysates. Theanti-933W toxin neutralized cytotoxic activity in culturesupernatants, but neither anti-Shiga toxin nor monoclonalantibody to Shiga-like toxin had this effect. That the 933Jtoxin was in fact present in culture supernatants was evidentfrom the more dramatic reduction in cytotoxic activity whena mixture of anti-Shiga toxin and anti-933W toxin was usedrather than anti-933W toxin alone.

Neutralization of the cytotoxic activity from clinical isolatesof E. coli. Marques et al. (L. R. M. Marques, M. A. Moore,J. G. Wells, I. K. Wachsmuth, and A. D. O'Brien, J. Infect.Dis., in press) recently surveyed 418 E. coli strains forproduction of Shiga-like toxin and identified 49 elevatedcytotoxin-producing isolates (moderate and high producers).The cell-associated cytotoxicity of 39 of these strains wasneutralized by anti-Shiga toxin, but for 10 strains the cell-associated cytotoxicity could not be neutralized by anti-Shiga toxin. In the present study, the antiserum prepared foranalysis of the 933W-encoded toxin was used (on the basis ofthe neutralization data in Table 2) to determine whether thecell-associated cytotoxicity of the 10 strains not neutralizedby anti-Shiga toxin could be neutralized by anti-933W crudetoxin. The isolates that produced toxin which was notneutralized by anti-Shiga toxin included four strains(0157:H7) from patients with hemorrhagic colitis, two

strains (0113:H21) from patients with hemolytic uremicsyndrome, one strain (0157:H7) from a patient with diar-rhea, and three strains (04:NM, 08:H9, 0149:H8) fromcalves with diarrhea. The cytotoxic activity in lysates fromall 10 isolates was neutralized by the rabbit antiserum againstcrude 933W toxin. In addition, the cell-associated cytotoxicactivity of one E. coli 0157:H- strain (E32511) described byScotland et al. (S. M. Scotland, H. R. Smith, and B. Rowe,Letter, Lancet i:885-886, 1985) was only neutralized byrabbit anti-933W toxin. The cytotoxic activity in culturesupernatants of E32511 and of 27 strains reported byMarques et al. (Marques et al., in press) as not neutralizedby anti-Shiga toxin, was found in this investigation to beneutralized by either anti-Shiga toxin plus anti-933W toxin oranti-933W toxin alone (the 10 strains described above).Therefore, some clinical isolates produce cytotoxin antigen-ically related to the 933W toxin, and this antigenic variant ofShiga-like toxin may play a role in the pathogenesis ofdiarrhea, hemorrhagic colitis, and hemolytic uremic syn-drome. The relative contributions of the 933J and 933Wtypes of Shiga-like toxin to the virulence of E. coli strainsthat produce both toxins remains to be determined.Homology of the 9331 and 933W toxin genes. To compare

the relatedness of the toxin genes at the DNA level, South-ern hybridization experiments were performed between re-striction fragments of phage 933W DNA and the pJN25 DNAprobe for the intact structural genes for Shiga-like toxin fromphage 933J. Under the more stringent hybridization condi-tions tested, estimated to permit no more than 26% base-pairmismatch, pJN25 hybridized to at least one band from phage933W DNA digested with EcoRI, SalI, or BamHI, eithersingly or in pairwise combinations (Fig. 1). In control ex-periments, labeled DNA from the pBR328 vector did nothybridize to the restricted 933W phage DNA under lessstringent conditions, estimated to permit up to 36% mis-match (data not shown). These results demonstrate that thecloned DNA insert is responsible for the hybridization of thepJN25 probe with the 933W DNA and provide evidence thatthe 933J and 933W toxins have homologous nucleotidesequences. Definitive proof for their homology, however,will require the use of probes that are known to be com-pletely internal to the toxin structural genes.

DISCUSSIONThe major conclusion from this study is that the Shiga-like

toxin encoded by phage 933W is antigenically distinct from,but homologous with, the Shiga-like toxin encoded by phage933J. Data from the cross-neutralization experiments withthe polyclonal antisera and with MAb 13C4, which is specificfor an epitope on the B subunit of Shiga-like toxin (25), didnot demonstrate any shared epitopes between the 933W and933J toxins. Nevertheless, the 933J- and 933W-encodedtoxins appear to be homologous both functionally and genet-ically. Their similar biologic activities, as well as theiridentical target cell specificities (Table 1), suggest that boththeir receptor-binding function and their mode of action,which in the case of Shiga toxin is inhibition of proteinsynthesis (22), may be similar. One difference in the biologicactivities of the 933W and 933J toxins is the ratio of LD50 toCD50 per milligram of cell lysate protein. The genetic relat-edness of the 933W and 933J toxins is suggested by thehomology between the cloned 933J toxin genes and 933Wphage DNA in the Southern hybridization studies.Evidence supporting the existence of antigenically distinct

cytotoxins from E. coli has been reported by several inves-tigators (7, 10, 11, 24, 27; Scotland et al., Lancet i:885-886,

INFECT. IMMUN.

on June 9, 2018 by guesthttp://iai.asm

.org/D

ownloaded from

ANTIGENIC VARIATION AMONG SHIGA-LIKE TOXINS OF E. COLI 139

933W4 DNA_ z

0E . -

FX X X COc

o o o- E j 0

wwwX en m e w

I,

ML ml

31-4 23.0

-.4 9.4-4 6.7

-q 4.4

--q 2.3-.4 2.04.

FIG. 1. Southern hybridization of933W phage DNA probed with32P-labeled pJN25, a plasmid containing the structural genes forShiga-like toxin from phage 933J. Hybridization of pJN25 wasperformed in an aqueous solution using 5 x SSPE overnight at 65°C.The autoradiograph represents a 4-day exposure in the presence ofan intensifying screen. Fragments of lambda DNA cut with Hindlll(Bethesda Research Laboratories) were used as molecular-sizemarkers. The multiple bands observed in some of the lanes indicatethat the restriction enzyme(s) cut one or more times within the DNAsequences which show homology with the probe.

Scotland et al. (Scotland et al., Lancet i:885-886, 1985)have suggested the names verotoxin 1 (VT1) for the toxinencoded by genes which hybridize with S. dysenteriae type1 and E. coli H19 (026:H11) and verotoxin 2 (VT2) for anantigenically distinct toxin produced by E. coli E32511(0157:H-). It is now clear that Shiga-like toxin I and VT1are different names applied to the same serological group oftoxins (13; O'Brien et al., Lancet i:702, 1983; Scotland et al.,Lancet i:885-886, 1985). The name verotoxin has been usedin reference to the Vero cell cytotoxin from E. coli that wasoriginally described in 1977 by Konowalchuk et al. (11).Thus, the term verotoxin was coined before the recognitionthat the Vero cell- and HeLa cell-active toxins produced byE. coli and S. dysenteriae 1 are antigenically related (25;O'Brien et al., Lancet i:702, 1983). We prefer the nameShiga-like toxin I to VT1 for the following reasons. (i) Itacknowledges the precedence of the name Shiga toxin (1; M.Neisser, and K. Shiga, Letter, Dtsch. Med Wochenschr.29:61, 1903) for toxins with the biological activities summa-rized in Table 1. (ii) It emphasizes the antigenic and geneticrelationships of the toxin with Shiga toxin. (iii) Cytotoxicityfor Vero cells is a property of many cytotoxins that is notunique to Shiga-like toxin. Shiga-like toxin II and VT2appear to be the same toxin as determined by neutralizationof E. coli E32511 cytotoxic activity with rabbit anti-933Wtoxin. It is not yet established whether Shiga-like toxin II isproduced by bacteria other than E. coli. The definitiverelationships between the members of the family of Shiga-like toxins will ultimately be defined at the genetic level bythe nucleotide sequences of their structural genes. Suchinformation will also provide a definitive basis for establish-ing a nomenclature that accurately reflects the relationshipsbetween Shiga toxin and the Shiga-like toxins.

ACKNOWLEDGMENTS

The work was funded by Public Health Service grant Al 20148-03from the National Institutes of Health and by grant no. 2210-00 fromthe Agency for International Development.We thank Samuel B. Formal and Hugh Collins of the Walter Reed

Army Institute of Research for performing the enterotoxicity as-says. We also thank Edda Twiddy, Uniformed Services Universityof the Health Sciences, for assisting with the Y-1 cytotoxicityassays. We gratefully acknowledge Sylvia Scotland, Sara Rothman,Helge Karch, Sherwood Gorbach, and Michael Doyle for commu-nicating to us their observations that. anti-Shiga toxin did notcompletely neutralize the cytotoxic activity present in culture su-pernatants of E. coli 933.We also thank Helge Karch for sharing hisunpublished data comparing the mouse lethality of E. coli strainsC600(933W) and C600(933J).

1985). To avoid confusion, the nomenclature for the anti-genic variants should be addressed. We propose that thename Shiga-like toxin I be applied to the 933J toxin andrelated toxins which are neutralized by antiserum againstShiga toxin from S. dysenteriae type 1 strain 60R. The nameShiga-like toxin II is suggested for the 933W toxin and toxinsneutralized by antibodies to 933W toxin. Specific antiseraraised against each purified prototype toxin should be usedto identify which antigenic variant is produced. We havemade antibodies against purified Shiga toxin and monoclonalantibodies against Shiga-like toxin I from E. coli, and we arenow in the process of purifying the 933W toxin and raisingpolyclonal and monoclonal antibodies against it. Once puri-fied 933W toxin is available, we will also be able to comparedirectly the specific activities of the 933J and 933W toxins ineach of the available bioassays.

LITERATURE CITED

1. Conradi, H. 1903. Ueber losliche, durch aseptische Autolyseerhaltene Giftstoffe von Ruhr- und Typhus-Bazillen. Dtsch.Med. Wochenschr. 29:26-28.

2. Davis, R. W., D. Botstein, and J. R. Roth. 1980. A manual forgenetic engineering. Advanced bacterial genetics. Cold SpringHarbor Laboratory, Cold Spring Harbor, N.Y.

3. Formal, S. B., D. Kundel, H. Schneider, N. Kunev, and H.Sprinz. 1961. Studies with Vibrio cholerae in the ligated loop ofthe rabbit. Br. J. Exp. Pathol. 42:504-510.

4. Gentry, M. K., and J. M. Dalrymple. 1980. Quantitative micro-titer cytotoxicity assay for Shigella toxin. J. Clin. Microbiol.12:361-366.

5. Holmes, R. K., and E. M. Twiddy. 1983. Characterization ofmonoclonal antibodies that react with unique and cross-reactingdeterminants of cholera enterotoxin and its subunits. Infect.Immun. 42:914-923.

VOL. 53, 1986

on June 9, 2018 by guesthttp://iai.asm

.org/D

ownloaded from

140 STROCKBINE ET AL.

6. Karmali, M. A., M. Petric, C. Lim, P. C. Fleming, G. S. Arbus,and H. Lior. 1985. The association between idiopathic hemo-lytic uremic syndrome and infection by verotoxin-producingEscherichia coli. J. Infect. Dis. 151:775-782.

7. Kashiwazaki, M., T. Ogawa, K. Nakamura, T. Isayama, K.Tamura, and R. Sakazaki. 1981. Vero cytotoxin produced byEscherichia coli strains of animal origin. Nat. Inst. Anim.Health Q. 21:68-72.

8. Keusch, G. T., A. Donohue-Rolfe, and M. Jacewicz. 1982.Shigella toxin(s): description and role in diarrhea and dysentery.Pharmacol. Ther. 15:403-438.

9. Keusch, G. T., and M. Jacewicz. 1977. Pathogenesis of shigelladiarrhea. VII. Evidence for a cell membrane toxin receptorinvolving (3 1--4-linked N-acetyl-D-glucosamine oligomers. J.Exp. Med. 146:535-546.

10. Konowalchuk, J., N. Dickie, S. Stavric, and J. I. Speirs. 1978.Comparative studies of five heat-labile toxic products of Esch-erichia coli. Infect. Immun. 22:644-648.

11. Konowalchuk, J., J. I. Speirs, and S. Stavric. 1977. Veroresponse to a cytotoxin of Escherichia coli. Infect. Immun.18:775-779.

12. Maniatis, T., E. F. Fritsch, and J. Sambrook. 1982. Molecularcloning: a laboratory manual. Cold Spring Harbor, Cold SpringHarbor Laboratory, N.Y.

13. Newland, J. W., N. A. Strockbine, S. F. Miller, A. D. O'Brien,and R. K. Holmes. 1985. Cloning of Shiga-like toxin structuralgenes from a toxin converting phage of Escherichia coli. Sci-ence 230:179-181.

14. O'Brien, A. D., and G. D. LaVeck. 1982. Immunochemical andcytotoxic activities of Shigella dysenteriae 1 (Shiga) and Shiga-like toxin. Infect. Immun. 35:1151-1154.

15. O'Brien, A. D., G. D. LaVeck, D. E. Griffin, and M. R.Thompson. 1980. Characterization of Shigella dysenteriae 1(Shiga) toxin purified by anti-Shiga toxin affinity chromatogra-phy. Infect. Immun. 30:170-179.

16. O'Brien, A. D., G. D. LaVeck, M. R. Thompson, and S. B.Formal. 1982. Production of Shigella dysenteriae type 1-likecytotoxin by Escherichia coli. J. Infect. Dis. 146:763-769.

17. O'Brien, A. D., J. W. Newland, S. F. Miller, R. K. Holmes,H. W. Smith, and S. B. Formal. 1984. Shiga-like toxin-converting phages from Escherichia coli strains that cause

hemorrhagic colitis or infantile diarrhea. Science 226:694-696.18. O'Brien, A. D., M. R. Thompson, P. Gemski, B. P. Doctor, and

S. B. Formal. 1977. Biological properties of Shigellaflexneri 2Atoxin and its serological relationship to Shigella dysenteriae 1toxin. Infect. Immun. 15:796-798.

19. Olsnes, S., and K. Eiklid. 1980. Isolation and characterization ofShigella shigae cytotoxin. J. Biol. Chem. 255:284-289.

20. Pai, C. H., R. Gordon, H. V. Sims, and L. E. Bryan. 1984.Sporadic cases of hemorrhagic colitis associated with Esche-richia coli 0157:H7. Clinical, epidemiologic, and bacteriologicfeatures. Ann. Intern. Med. 101:738-742.

21. Reed, L. J., and H. Muench. 1938. A simple method of estimat-ing fifty percent end points. Am. J. Hyg. 27:493-497.

22. Reisbig, R., S. Olsnes, and K. Eiklid. 1981. The cytotoxicactivity of Shigella toxin. Evidence for catalytic inactivation ofthe 60S ribosomal subunit. J. Biol. Chem. 256:8739-8744.

23. Riley, L. W., R. S. Remis, S. D. Helgerson, H. B. McGee, J. G.Wells, B. R. Davis, R. J. Hebert, E. S. Olcott, L. M. Johnson,N. T. Hargrett, P. A. Blake, and M. L. Cohen. 1983. Hemor-rhagic colitis associated with a rare Escherichia coli serotype.New Engl. J. Med. 308:681-685.

24. Smith, H. W., P. Green, and Z. Parsell. 1983. Vero cell toxins inEscherichia coli and related bacteria: transfer by phage andconjugation and toxic action in laboratory animals, chickens,and pigs. J. Gen. Microbiol. 129:3121-3137.

25. Strockbine, N. A., L. R. M. Marques, R. K. Holmes, and A. D.O'Brien. 1985. Characterization of monoclonal antibodiesagainst Shiga-like toxin from Escherichia coli. Infect. Immun.50:695-700.

26. Wells, J. G., B. R. Davis, I. K. Wachsmuth, L. W. Riley, R. S.Remis, R. Sokolow, and G. K. Morris. 1983. Laboratory inves-tigation of hemorrhagic colitis outbreaks associated with a rareEscherichia coli serotype. J. Clin. Microbiol. 18:512-520.

27. Willshaw, G. A., H. R. Smith, S. M. Scotland, and B. Rowe.1985. Cloning of genes determining the production of verocytotoxin by Escherichia coli. J. Gen. Microbiol. 131:3047-3053.

28. Wilson, G., and G. Smith. 1984. Diphtheria and other diseasesdue to corynebacteria, p. 91-94. In G. R. Smith (ed.), Topleyand Wilson's principles of bacteriology, virology, and immunol-ogy, vol. 3. The Williams & Wilkins Co., Baltimore.

INFECT. IMMUN.

on June 9, 2018 by guesthttp://iai.asm

.org/D

ownloaded from